4 Ways AI is Uncovering Secrets of the Past
Updated: Jul 23, 2022
Artificial Intelligence is no longer a fresh curiosity we know little about. Its applications have branched into almost all industries imaginable: fashion, transport, marketing, entertainment, healthcare, etc.
It is also being increasingly used in the scholastic world, especially among people studying our past. From a time in the 2000s, when only about six scientific papers referred to the use of AI, more than 65 papers were published in 2019 talking about its archaeological applications alone.
Linguists, anthropologists, historians and archaeologists are using techniques like deep learning and convolutional neural networks (CNNs) to unearth ancient settlements, decipher mysterious tongues, and also to find traces of new, unknown ancestors in our DNA.
In this post, we’ll look at 4 ways AI has been uncovering secrets of the past, and how it has revolutionized the study of history.
Why is AI Important?
So why has AI become such a popular tool to explore our past?
The reasons range from climate change to time-consuming traditional methods. Artificial Intelligence can sift through inputs far faster than humans can. Since the traditional methods take time and are labor intensive, it makes much more sense to employ artificial intelligence to do the job. This also makes them crucial in the race against time to identify archaeological sites that might face the wrath of global warming soon.
Software like the MyHeritage AI tool can make old historical photographs smile and frown, making titans of old human and approachable perhaps for the first time in history. So apart from being quick and affordable, AI has helped history become the most intimate and relatable it has ever been.
1. Bringing heroes (and villains) of history to life
In 2021, D-ID, an AI Face Technology company, in collaboration with MyHeritage, a genealogy site, rolled out Deep Nostalgia, an AI software that brings to life old photographs users upload to it. The AI achieves this by first enhancing and then stitching the enhanced picture to a driver video which contorts the portrait into different emotions.
People have been uploading old photographs of their ancestors to the site to watch them come to life. Animated faces of historical figures also took the internet by storm. People shared photos of Rosalind Franklin, Alan Turning, Hitler, Mark Twain among others blinking, smiling, twisting their heads. MyHeritage also used this technology to create an ad starring Abraham Lincoln.
Apart from animating old historical photographs, AI is also being used to guess what pre-camera figures might have looked like. Nathan Shipley, a VFX artist, makes realistic images of historical figures based on their portraits and busts. He uses machine learning and other AI techniques like general adversarial networks (GAN) to arrive at his apparently accurate guesses. His work on realistic images ranges from Henry VIII to Benjamin Franklin. As technical and co-creative director at the Dali Museum, Shipley also deepfaked Salvador Dali to make him and his works more personable to visitors.
2. Deciphering lost tongues
AI is also piecing together mysterious languages. Aimed at discovering relationships between ‘lost’ tongues, MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has launched an AI that can decipher such languages by itself, without needing any extra knowledge of their relationships with other languages.
The ‘decipherment algorithm’ classifies words from one ancient language and associates them with similar words in other related tongues. While the software is not a proper ‘translator’, it can identify the languages’ roots. The algorithm, for example, was able to correctly identify the Iberian language family, concluding that the language was not related to Basque, as recent research has confirmed.
Similarly, Google’s PYTHIA is a ‘restoration model that recovers missing characters from a damaged text input using deep neural networks,’ according to its research published in ARXIV, a Cornell University journal. It analyzes a series of damaged ancient text and then predicts hypothesised sequences with a character error rate of 30.1%, a drastic drop from the 57.3% error rate of traditional historians.
3. Finding archaeological sites
Archaeologists are also using AI to find new potential sites. While traditional methods take too long to go through and glean meaningful inputs from a deluge of aerial and satellite data, AI can do so far more quickly and efficiently.
Dylan Davis, a PhD candidate studying archaeological sites in Madagascar and South Carolina, has created such algorithms to locate new sites. The predictive algorithm he used in Madagascar helped his team identify upwards of 70 new sites from an area spanning more than a 1000 square kilometers in about a year, SingularityHub reported.
AI is also helping map ancient civilizations. The Institut Català d’Arqueologia Clàssica (ICAC) used machine learning algorithms to reconstruct 'more than 20,000 kilometers of paleo-rivers around the Indus Valley civilisation', according to Analytics Insight.
AI’s capacity to identify archaeology from the data provided is also being used to protect such sites from damage during construction. Iris Kramer’s ArchAI does this by using machine learning to detect archaeology during the early planning period itself, so that one may calculate beforehand the time and cost involved in a construction project, while also remarkably reducing the risk of finding unexpected archaeology at the construction site.
4. Extracting unknown ‘ghosts’ from our DNA
AI is also quickly becoming a necessity in genomics. The completion of the draft human genome sequence 20 years ago has since led to the creation of a baffling quantity of genomic data. The National Human Genome Research Institute says that within the next decade, genomics research will create an estimated 2 to 40 exabytes of data (one exabyte is equal to one quintillion bytes).
Since DNA sequencing and accompanying techniques will keep increasing the amount and complexity of such data sets, AI and ML based tools will be needed to sift through, categorize and glean information from such a vast body of data.
While AI seems to shape the future of genomics, identifying genetic disorders, predicting the spread of cancer in patients and bettering the performance of gene editing tools like CRISPR, it has also found another curious function in the field: looking into the past.
In 2019, while analyzing the human genome with the help of AI, scientists discovered an unknown species of human ancestors. The deep learning algorithms parsing through ancient and modern genetic code found the remains of this species lurking in Eurasian DNA, Science Alert reported. Scientists believe this to be the evidence of a “third introgression”: a ‘ghost’ species human beings mated with during the Out of Africa event.
This species was possibly a mix of the other two ancestor species known to us: the Neanderthals and the Denisovans, and while an earlier discovery of a hybrid fossil at Denisova seems to strongly suggest so, other possibilities cannot be done away with yet, the researchers say.
AI’s time-saving abilities have helped archaeologists and genomicists achieve in years what would have otherwise taken decades. Its inventive use to make historical figures more approachable than ever has given us a new appreciation for history. And while AI is making significant strides in the study of history, it does not mean that it is going to surpass the keen eye and judgment of human beings anytime soon.